S

GC

a

ARR1AA

KFFFMCP

1

aicfitrfmemtacvtie

l

0h

Journal of Chromatography A, 1324 (2014) 215– 223

Contents lists available at ScienceDirect

Journal of Chromatography A

jou rn al hom epage: www.elsev ier .com/ locate /chroma

treamlined analysis of lactose-free dairy products

ertrud E. Morlock ∗, Lauritz P. Morlock, Carot Lemohair of Food Science, Institute of Nutritional Science, Justus Liebig University Giessen, Heinrich-Buff-Ring 26-32, 35392 Giessen, Germany

r t i c l e i n f o

rticle history:eceived 1 October 2013eceived in revised form9 November 2013ccepted 20 November 2013vailable online 26 November 2013

eywords:unctional foodood safety

a b s t r a c t

Functional food for lactose-intolerant consumers and its global prevalence has created a large market forcommercially available lactose-free food products. The simplest approach for detection and quantitationof lactose in lactose-free dairy products was developed. A one-step sample preparation was employedand the resulting 10% sample solution was directly subjected to the chromatographic system. LODs downto 0.04 mg/L were obtained for dairy products by application volumes up to 250 �L on a rectangular startzone, which is the lowest LOD reported in matrix so far. The highly matrix-robust, streamlined approachwas demonstrated for a broad range of dairy products, even with high fat and protein contents. The meanrecovery rate for 11 types of dairy products spiked at the strictest lactose content discussed (0.01%) was90.5 ± 10.5% (n = 11). The mean repeatability for 11 dairy products spiked at the 0.01% level was 1.3 ± 1.0%

ood quality controlilk products

arbohydrateslanar chromatography

(n = 11). It is the simplest approach with regard to sample preparation at low running costs (0.3 Euro or 0.4USD/analysis) and fast analysis time (3 min/analysis). It enabled an efficient product screening, and at thesame time, the quantitation of lactose in relevant samples. This streamlined analysis is highly attractiveto the field of food safety and quality control of lactose-free dairy products, for which a limit value forlactose is expected soon in the EU. This methodological concept can be transferred to other challengingfields.

. Introduction

The triple A in food industry, i.e. availability all over the world atny place and time, added value (functionality), and affordability,s still the pacesetter [1]. Functional food for lactose-intolerantonsumer and its global prevalence has created a large marketor commercially available lactose-free food products. Lactose-ntolerant individuals have a deficiency of the enzyme lactase, andhus, lactose is not completely catabolized into its monosaccha-ide units glucose and galactose. Lactose is the major disaccharideound in milk and milk products. The various animal milks contain

ostly up to 5% lactose, but for example horse and ass milk containven up to 7% lactose [2]. ‘Lactose-free’ milk and milk products areostly produced by breaking down lactose into glucose and galac-

ose by enzymatic hydrolysis with ß-galactosidases (often labelleds lactase on the food). However, the resulting milk products mightontain varying amounts of residual lactose. Lactose intolerancearies widely among individuals with lactose maldigestion and the

hreshold of lactose is highly individual [3]. Additionally, especiallyn the newborn screening, the analysis of galactose is of inter-st as some individuals suffer from the genetic metabolic disorder

∗ Corresponding author. Tel.: +49 6419939141; fax: +496419939149.E-mail addresses: Gertrud.Morlock@ernaehrung.uni-giessen.de (G.E. Morlock),

auritz8997@web.de (L.P. Morlock), c lemo@hotmail.com (C. Lemo).

021-9673/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.chroma.2013.11.038

© 2013 Elsevier B.V. All rights reserved.

galactosemia. Such individuals do not tolerate lactose and addition-ally galactose. For example, ‘lactose-free’ milk beverages, in whichlactose is enzymatically hydrolyzed to glucose and galactose andfrom which the latter is not subsequently removed, are not suitablefor patients with galactosaemia.

Hence, it is discussed in Europe, how the lactose content oflactose-free food products will be defined. For example, the workinggroup Issues of Nutrition of the German Society of Food Chemistry(LChG) has recommended three categories of food declaration [4],i.e. the low in lactose level for food products of ≤1% lactose, very lowin lactose level (≤0.1%) and lactose-free level (≤0.01% of lactose andits degradation products). The latter is the strictest level being dis-cussed, which permits that these food products can safely be usedin the dietetic management of patients even with galactosaemia, assuch food products are also indicated to be ‘free’ of galactose. Fromthe point of view of nutrition and of consumers with different lac-tose thresholds, this 3-level categorization is rational for findingadequate food products on the market.

However, the German milk industry has a different opinionabout lactose-free food products and proposes levels ≤0.1% [5],which equals to the very low in lactose level discussed before andis allowed to contain lactose degradation products. The production

of lactose-reduced milk by hydrolysis of lactose is harmonized andregulated by EU law [6], whereas the production of other lactose-reduced products adhere to national law, which regulates that adairy is only allowed to add the ß-galactosidase on the authority

2 matog

ofaTgwbfipptotcoppmtspsr

lpfsebilvltllfhad0eaacttsfsitfl

2

2

siaaS

16 G.E. Morlock et al. / J. Chro

f the respective German ministry (BMELV) [7]. The final lactose-ree product tastes sweeter if compared to the standard products the breakdown products are more intense in their sweetness.he production of lactose-free products ≤0.01% of lactose, but notalactose-free, is patented and the application of the technologyould require a license [8]. This technology involves a 3-step

ased membrane filtration followed by re-combination: first, ultra-ltration (UF) of the milk, secondly nanofiltration (NF) of the UFermeate, thirdly reverse osmose (RO) for concentration of the NFermeate to a salt, and fourthly rearrangement, i.e. combination ofhe RO and UF retentate. By doing so, half the lactose is separatedut of the milk without any other changes to the milk’s composi-ion. The subsequent addition of lactase splits the remaining lactoseontent into glucose and galactose. The final product gains the samerganoleptic characteristics as standard milk and completely com-lies with consumer expectations. With regard to galactosaemiaatients, the further absence of galactose would require additionalilk treatment steps. Thus, the German milk industry advocacy of

he ≤0.1% lactose level is understandable as the production of thetrictest lactose-free level (≤0.01% of lactose and its degradationroducts according to the LChG working group) generates expen-ive extra costs and might not have an adequate high consumptionate.

It remains exciting how the decision of the EU commission aboutactose limit values will be made in the near future. Not only for foodroduction the strictest lactose-free level is challenging, but alsoor food analysis. Many currently available analytical methods forugar analysis, like polarimetry, mid-infrared detection, photom-try/fluorometry, and gravimetry, do not allow the differentiationetween carbohydrates and are not suitable for measuring lactose

n food products with diverse sugars. Enzymatic assays [9–12] andiquid column chromatography in combination with diverse uni-ersal detectors, e.g., refractive index detector (RID), evaporativeight-scattering detector (ELSD), and corona charged aerosol detec-or (CAD), miss the required capability of detection for the strictestevel of 0.01% discussed [13]. For example, using HPLC–CAD, theimit of quantitation (LOQ) of lactose in lactose-reduced low-at milk was estimated to be 0.02% [14]. However, recently aigh-performance anion-exchange chromatography with pulsedmperometric detection was reported, for which the method’setection limit (LOD) for the lactose standard solution was.12 mg/L (0.000012%) using a 65-min gradient [15]. Also a capillarylectrophoresis method with electrochemical detection reported

LOD of 0.1 mg/L (0.00001%) for the lactose standard solution at separation time of 24 min [16]. Such methods would allow theontrol of the strictest level discussed (100 mg/L or 0.01%). Never-heless, food control demands fast, cost-effective and at the sameime matrix-robust analytical methods with regard to the varyingample matrices of the meanwhile large assortment for ‘lactose-ree’ products. The following study describes, to our knowledge, theimplest approach for accurate and robust determination of lactosen ‘lactose-free’ dairy products with a one-step sample prepara-ion followed by high-performance thin-layer chromatography anduorescence detection (HPTLC–FLD) after selective derivatization.

. Materials and methods

.1. Materials

d(−)-fructose, d(+)-maltose-1-hydrat, d(+)-mannose anducrose (all ≥ 98%) as well as acetonitrile (≥99%), acetone, butanol,

-propanol, methanol, i-propylacetate and ethyl acetate (all ≥ 99%)s well as aniline (≥99.9%), diphenylamine (≥98%), p-aminobenzoiccid (≥99%) and o-phosphoric acid (85%) were obtained from Flukaigma Aldrich, Seelze, Germany. d(+)-lactose-1-hydrat (Ph. Eur.)

r. A 1324 (2014) 215– 223

were delivered by Roth, Karlsruhe, and d(+)-glucose and d(+)-galactose (GPR Rectapur) by vwr, Darmstadt, Germany. Acetic acid(≥99.8%), sulphuric acid (98%), boric acid (≥99.9%), 2-naphthol,ninhydrin (both per analysis) and HPTLC plates silica gel 60,20 cm × 10 cm, were provided by Merck, Darmstadt, Germany.Alternatively, plates with indicator F254 can be used. Distilledwater was produced by Heraeus Destamat Bi 18 E (Thermo FisherScientific, Schwerte, Germany) and deionized water by a SynergySystem (Millipore, Schwalbach, Germany). All food samples werepurchased at local stores.

2.2. Extraction of ‘lactose-free’ food samples

2.5 g of each sample were dissolved with 8 mL distilled or deion-ized water in a 25-mL volumetric flask. For cheese and chocolatesamples, 70 ◦C hot water was used. The sample was stirred for10 min with a magnetic stir bar on the magnetic stirrer. Afterremoval of the stir bar by rinsing with methanol (and coolingdown), the flask was filled up to the mark with methanol. Thisresulted in a final concentration of 100 mg/mL in methanol–water2:1, v/v. An aliquot of each sample was centrifuged (13.000 g, 5 min)and stored at 4 ◦C until analysis.

2.3. Standard (mixture) solutions

For mobile phase optimization, two methanolic sugar mixturesolutions were prepared (100 ng/�L each), which contained fruc-tose, galactose and glucose (mix 1) and lactose, maltose, sucroseand mannose (mix 2). For analysis of dairy products, aqueous stocksolutions of lactose, glucose and galactose (100 ng/�L each) werediluted 1:33 together in mixture 3 (mix 3; 3 ng/�L) and 1:100 inmixture 4 (mix 4; 1 ng/�L) using methanol. For spiking of dairyproducts, an aqueous lactose solution (3 �g/�L) was used and its1:300 methanolic dilution for calibration (10 ng/�L).

2.4. Application

The solutions were sprayed-on as 8, 11 or 15 mm bands or15 mm × 3 mm rectangles with a track distance of 9, 12 or 16 mm,respectively, using the Automatic TLC Sampler 4 (ATS 4, CAMAG,Muttenz, Switzerland). The distance to the lower edge was setat 8 mm. The dosage speed was investigated as follows: 150 and250 nL/s (bands) as well as 250 to 1000 nL/s (rectangles) with andwithout heating nozzle at 60 ◦C. For calibration, 1, 5, 10, 15 �L(3–60 ng/band), 1, 10, 20, 30, 40 �L (3–120 ng/band) or 1, 30, 60,100 �L (3–300 ng/band) were sprayed-on using mix 3, or 1, 5, 10,15 �L (10–150 ng/band) using the diluted lactose solution. 1 to250 �L sample solution were applied depending on the applicationzone geometry (8, 11, 15 mm bands or 15 × 3 mm rectangles). Over-lapped application (overlap of 50%) was performed by applicationof 16 mm bands each of a sample (2.5 �L) and standard solution(50 �L of mix 3). After application the plate was dried for 30 s in astream of warm air, or automatically in the Automated DevelopingChamber ADC 2 (CAMAG).

2.5. Chromatography

Mobile phase development was carried out on cut plates(5 × 10 cm) in the Twin Trough Chamber (10 × 10 cm, CAMAG)with varying mobile phase combinations and ratios. Food analy-sis was performed using i-propyl acetate–methanol–water 11:7:2

(v/v/v/v) as mobile phase up to a migration distance of 60 mm. Theseparation took 30 min in the Twin Trough Chamber 20 × 10 cmor ADC 2. The relative air humidity was 40 ± 10% and the roomtemperature 23 ± 5 ◦C during the whole study.

matog

2

rIt(swatt2arawpw1i

2

TTia

2

S6TotAwt1

3

3

(fad(al

varu[t1ato

G.E. Morlock et al. / J. Chro

.6. Derivatization

After drying of the plate, it was automatically immersed into theespective derivatization reagent using the TLC Immersion DeviceII (CAMAG; vertical speed 4 cm/s; immersion time 0 s). For deriva-ization with diphenylamine aniline o-phosphoric acid reagentDAP), a mixture of 70 mL aniline solution, 70 mL diphenylamineolution (2% each in acetone) and 10 mL o-phosphoric acid (85%)as prepared. For the p-aminobenzoic acid reagent (ABA), 1 g p-

minobenzoic acid was dissolved in 36 mL pure acetic acid, andhen, 40 mL water, 2 mL o-phosphoric acid (85%) and 120 mL ace-one were added. For the 2-naphthol sulfuric acid reagent (NSA),

g 2-naphthol were dissolved in 180 mL ethanol and 12 mL sulfuriccid (50%) were added dropwise to avoid an accelerated exothermaleaction. For the o-phthalaldehyde reagent (OPA), 1.5 g o-phthaliccid and 1 mL aniline were dissolved in 100 mL n-butanol saturatedith water. A 0.1% methanolic ninhydrin solution was also pre-ared. After immersion into the derivatization reagent, the plateas heated on the TLC Plate Heater III (CAMAG) at 110 ◦C (ABA

40 ◦C) for 5 min, and for OPA 130 ◦C for 15 min. All reagents storedn the refrigerator were stable for at least two months.

.7. Documentation

Plate images captured at a gain of 1 were documented by theLC Visualizer or DigiStore2 Documentation System (both CAMAG).he exposure times used were mostly around 30 ms (white lightllumination in the combined reflectance and transmission mode)nd 1500 ms (UV 366 nm, reflectance mode).

.8. Densitometry

Diffuse reflectance measurement was performed with the TLCcanner 3 or 4 (both CAMAG). The measurement slit dimension was, 8 or 12 mm × 0.40 mm for 8, 11 or 15 mm bands, respectively.he scanning speed was 20 mm/s. After derivatization with DAPr NSA, the absorbance was measured at 370 and 500 nm, respec-ively, using a combined deuterium and halogen tungsten lamp.fter derivatization with ABA or OPA, fluorescence measurementas performed at 366/>400 nm using the mercury lamp. Quanti-

ative evaluation was performed with winCATS software version.46.

. Results and discussion

.1. Selection of mobile phase

Different solvent systems were investigated as mobile phaseall listed as volume ratios). Solvents of a low viscosity were pre-erred to enable a fast separation of the sugars. However, the overallim was to generate a wide separation distance of lactose to itsegradation products galactose and glucose as well as to sucrosemain sugar in food products). A wide separation distance to lactosellowed a high matrix load onto the plate, and thus low detectionimits.

Mobile phase mixtures for sugars were mostly based on highiscous solvents like butanol, e.g., n-butanol–i-propanol–aceticcid–2% aqueous boric acid 6:14:1:3, [17], for which the sepa-ation took approximately 1 h. Thus, fast mobile phase mixturessed for separation of polar analytes in other application fields18–20] were transferred and adjusted to lactose analysis. Mix-ures of ethyl acetate–methanol–water 7:2:1 or (slightly less polar)

4:5:1 as well as acetonitrile–water 4:1 or 9:1 did not yield

satisfying resolution with regard to sugar mixture 1. In mix-ures of ethyl acetate–2% aqueous boric acid–formic acid 14:5:3r ethyl acetate–water–formic acid 73:22:15, sucrose was located

r. A 1324 (2014) 215– 223 217

near to lactose. Often boric acid derivatives were added to mobilephase mixtures for sugar separation, which was shown to improvethe resolution between glucose and fructose, e.g., 2-aminoethyldiphenylborate in concentrations of 1.5–3 mmol/L [21]. However,the addition of boric acid derivatives was not helpful for the res-olution with regard to lactose. Finally, the mobile phase systemconsisting of i-propyl acetate–methanol–water 11:7:2 was foundto be a good choice and employed for all subsequent separations. Itwas fast and the resolution of glucose (hRF 43) and galactose (hRF

53) to lactose (hRF 28) was highly satisfying (R > 1.5; calculated for3–120 ng/band). The development up to 60 mm took 30 min in par-allel for 20 analyses. The elution order was determined by use ofthe single sugar solutions (not shown).

3.2. Investigation of detectability

For selective and sensitive detection of lactose and its degra-dation products (galactose and glucose) in dairy products, fourdifferent derivatization reagents were compared. LOD (S/N 3) aswell as LOQ (S/N 10) were determined for lactose. For NSA andDPA, the visual detectability was monitored and the plate was takenfrom the TLC Plate Heater after 5 min. Longer heating times led to anincreased colorization of the plate background. For OPA, the heat-ing at 130 ◦C for 15 min (investigated between 5 and 15 min) wasby a factor of 2.5 improved in contrast to 100 ◦C for 5 min. For ABA,a detailed study followed as it was found to perform best in thecomparison due to the selective and sensitive derivatization of lac-tose and its degradation products to a blue fluorescent derivative.For fluorescence detection with the mercury lamp, the excitationwavelength at 366 nm was superior to other main emission linesof the lamp like UV 254 or 313 nm. The signal intensity of lactoseincreased up to a temperature of 140 ◦C (for heating time of 5 min)and until a duration of 15 min, when heated at 110 ◦C (Fig. S-1). Fora fast protocol, heating at 140 ◦C for 5 min was selected for all sub-sequent ABA derivatizations. After one day, the signal intensity lost30%, when the plate was stored in the dark. Thus, a direct evaluationwas employed to guarantee a good detectability.

ABA and NSA were comparable in the detectability for lac-tose (LOD 1 ng/band and LOQ 3 ng/band, Table 1) and superiorwith regard to DAP (LOD 3 ng/band and LOQ 6 ng/band) and OPA(LOD 12 ng/band and LOQ 39 ng/band). With regard to all threecarbohydrates, NSA detected them at a comparable sensitivity;whereas galactose and glucose were detected up to a factor 13more sensitive than lactose using ABA (Table 1). Thus, ABA waspreferred to NSA and used in further experiments due to theimproved detectability of galactose. Nevertheless, NSA was com-parable with regard to lactose detection and analysis in complexfood matrix, and it was highly suited, too. Owed to the direct visu-alization of lactose contents as brown bands after derivatization,NSA was advantageous for rapid screening of samples in the visualrange.

3.3. Minimization of sample preparation

Dairy samples were prepared highly concentrated, as 10% sam-ple solutions (100 mg/mL, Table 2). First water was added todissolve the sugars, and then the solution was filled up withmethanol, which evaporated faster during spray-on application andenabled a faster dosage speed, if compared to solely water. Samplepreparation steps like solid phase extractions and/or Carrez I/II pre-cipitations were not necessary. By doing so, the sample preparation

and application of 1-�L samples was very fast and enabled an LODof 10 mg/L or 0.001% (Table 3), which was below the strictest lactoselevel discussed so far (100 mg/L). The 17 sample matrices analyzed,like cream, cheese, yoghurt etc. (Table 2), proved the robustness

218 G.E. Morlock et al. / J. Chromatogr. A 1324 (2014) 215– 223

Table 1Comparison of LOD and LOQ of four different derivatization reagents investigated for lactose detection.

Derivatization reagent ABA NSA DAP OPA

Detection mode Fluorescence Absorbance Absorbance FluorescenceWavelength (nm) 366/>400 500 370 366/>400Zone hue blue Red–brown Blue–grey BlueDetection via peak area/heightLOD (ng/band) [S/N] 1/1 [3/3] 1/1 [3/3] 3/3 [3/2] 12/12 [4/3]LOQ (ng/band) [S/N] 3/3 [10/9] 3/3 [10/9] 6/6 [10/7] 39/39 [12/8]Increased detection factor for glucose 13/5 1/1 2/2 4/2Increased detection factor for galactose 13/7 1/1 1/1 7/5

Table 2Analysis of lactose in 17 ‘lactose-free’ labelled dairy products (<0.1 g/100 g) commercially available in Germany down to the LOD level of 0.001% (1-�L sample volumesapplied, corrected by recovery rate), which allowed the control of the strictest LOD level discussed (0.01%).

Lactose-free dairy product No Dairy product name German manufacturer Lactose found (%) Repeatability(%RSD, n = 3)

Butter 1 Süßrahmbutter, lactose-free Schwarzwaldmilch,Freiburg

<0.001 –Yoghurt 2 Joghurt mild, cremig gerührt, lactose-free <0.001 –Milk 3 Frische fettarme Milch, lactose-free 0.007 8.3

4 H-Milch, lactose-free, fat reduced Elite/Penny, Köln <0.001 –Evaporated milk 5 Kaffeesahne, MinusL, lactose-free Omira, Ravensburg 0.006 10.1Buttermilk 6 Buttermilch-Drink, MinusL, lactose-free, max. 1% fat <0.001 –Sour cream 7 Sauerrahm, lactose-free, 10% fat Schwarzwaldmilch,

Freiburg<0.001 –

Cream 8 Schlagsahne, lactose-free 32% fat <0.001 –Goat cheese 9 Bio Ziegen-Butterkäse, Andechser Natur, Bioland Andechser Molkerei Scheitz, Andechs <0.001 –

10 Alpenmark Ziegenkäse Natur ALDI Süd, Mülheiman der Ruhr

0.064 10.011 Alpenmark Ziegenkäse Bockshornklee <0.001 –

Cheese 12 Alpenmark Kasländer Paprika-Jalapeno <0.001 –13 Alpenmark Kasländer Mild und Nussig <0.001 –14 Butterkäse Heirler Cenovis, Radolfzell <0.001 –

reese-fre

oo

3

swvpaanws(vlt

u

TL

15 Gemischter Käseaufschnitt MinusL, lactose-fCream cheese 16 Frischkäse, Doppelrahm- stufe, MinusL, lacto

17 Frischkäse, Magerstufe, lactose-free, 0.2% fat

f the method with regard to these varying matrices at the givenne-step sample preparation.

.4. LOD and LOQ with/without matrix for bandwise application

The LOD (1 ng/band) and LOQ (3 ng/band) of lactose standardolutions after derivatization with ABA (Table 1) were comparedith the limit values in various matrices of dairy products. The limit

alues with and without matrix (exemplarily proven for dairy sam-les no. 1-3, 5-9, 15-17 with and without overspray application of 1nd 3 ng/band lactose) were comparable. This proved that the sep-ration from matrix was good and background interferences wereot evident. The spike of 1 and 3 ng/band into the different matricesas visible in the chromatogram and densitogram, when 1 or 10 �L

ample volumes were applied to control the LOD of 10 or 1 mg/L0.001% or 0.0001%), respectively. However, for higher applicationolumes and thus lower LODs, all ‘lactose-free’ samples contained

actose and the original sample signal had to be subtracted fromhe spiked samples.

Depending on the matrix, different application volumesp to 80 �L were proven for a band length of 8 mm, used for

able 3OD of lactose in dairy products depending on matrix loading (increased application volu

Dairy product Cheese Butter Milk prod.

Dairy sample no. 9, 15 1 2,3,5–8

Band length or rectangle (mm) 8

Measurement slit (mm) 6

Sample solution (mg/mL) 100

Application volume (�L) 80 30 1

Sample amount applied (�g/band) 8000 3000 100

LOD in matrix (ng/band) 1

LOD (%) 0.00001 0.00003 0.001

LOD (mg/L) 0.1 0.3 10

Omira, Ravensburg <0.001 –e <0.001 –

Schwarzwaldmilch, Freiburg <0.001 –

simultaneous analysis of 20 tracks (Table 3). The LOD was loweredwith an increased application volume, exemplarily shown forbutter and cheese matrices (no. 1, 9 and 15). The 11-mm bandlength allowed the application of 100 and 50 �L volumes of cheeseand butter, respectively, on 15 tracks per plate. By doing so, LODwas lowered down to 0.00001%.

3.5. Lowering LOD in matrix through rectangular application

This potential for reduced LODs was investigated next for rect-angular applications of 15 × 3 mm. The band length of 15 mm waslimited due to the measurement slit dimension of 12 mm, whichwas the maximal slit length selectable in the software. For evalua-tion of bandwise applications, the homogenous middle part of theband (70–80% aliquot) was used for measurement. The minimumband width (3 mm) for rectangular application (in software called‘area’) was selected to avoid an extra focusing step and to keep

the protocol as simple as possible. Depending on the dairy sample,the acceptable matrix loading was different. Application volumesof 300 �L cheese samples (no. 9 and 15) were already overloaded.However, the following application range and resulting LOD level

mes required an increased band length or rectangular application).

Cheese Butter Cheese Butter Milk prod.

9, 15 1 9, 15 1 2

11 15 × 38 12100 100100 50 250 150 2010,000 5000 25,000 15,000 20001 10.00001 0.00002 0.000004 0.000007 0.000050.1 0.2 0.04 0.07 0.5

G.E. Morlock et al. / J. Chromatog

Fig. 1. Application volumes between 150 and 250 �L of cheese sample no. 9: lactosewti

wcattul

1tlptvaitats1L

mimiawtA

as detected in the cheese matrix with lowered hRF value at increasing applica-ion volumes (A) due to the increased fatty matrix loading visible at white lightllumination in the reflectance/transmission mode (B).

ere investigated and proved to work for the given sample matri-es (Table 3). For cheese samples, 80 to 250 �L applications werepplied to control the 0.00001 to 0.000004% lactose level. For but-er samples, 30 to 150 �L were investigated to control the 0.00003o 0.000007% lactose level. For milk and all other milk-based prod-cts, 1 to 20 �L were proven to control the 0.001 to 0.00005% lactose

evel.For increased application volumes of cheese samples between

50 and 250 �L (Fig. 1), the lactose zone was slightly deformed athe edges of the lactose band formed. This was caused by the non-inear mobile phase flow, which was slightly retarded in the middleart of the band due to the heavy matrix loading if compared tohe flow between the rectangles. Nevertheless, lactose was clearlyisible in the cheese matrix. Its hRF value was lowered at increasingpplication volumes (Fig. 1A). The lactose shift was caused by thencreased fatty matrix load visible at white light illumination ofhe plate image (Fig. 1B). Thus, 250 �L for cheese samples were stillcceptable and lowered LOD down to 0.04 mg/L (0.000004%). This ishe lowest LOD in matrix for control of ‘lactose-free’ food reportedo far. The LOD reduction was also demonstrated for butter (no. 1,50 �L) and another milk product (no. 2, yoghurt, 20 �L) with anOD of 0.000007% and 0.00005%, respectively.

To conclude, rectangular application coped with the highestatrix loading of 250 �L/rectangle cheese sample. The matrix load-

ng depended on the inherent sample matrix and the resultingaximal application volume, which lowered the LOD, but required

ncreased band lengths (11 mm instead of 8 mm) or even rect-

ngular application (15 × 3 mm, Table 3). If the high matrix loadas distributed over a longer band/rectangle, and at the same

ime, the measurement slit was increased, the LOD was improved.n increase of the band length to 11 or 15 mm limited the track

r. A 1324 (2014) 215– 223 219

number to 15 or 11 tracks per plate, respectively. In case of ‘lactose-free’ milk and milk-products, for which milk was enzymaticallydegraded, the degradation products glucose and galactose werestill present in the products. The excess of the degradation prod-ucts impaired the LOD of lactose, which was 0.00005% (Table 3).For other dairy products like butter and especially cheese, the LODof lactose was better owed to the production technology of theseproducts (substantially low in degradation products).

3.6. Dosage speed study

For application of higher sample volumes, the dosage speed wasadjusted to allow a rational application time. For bandwise appli-cation of ≥80-�L volumes, heating of the spray nozzle was used.If this option is not available, the dosage speed has to be reducedto 150 nL/s. For rectangular application, dosage speeds of 250, 300,400, 700 and 1000 nL/s with and without heating spray nozzle at60 ◦C were investigated (Fig. S-2). Lactose was visible on all tracks,although deformed due to the non-linear flow discussed before.Using the dosage speed of 700 or 1000 nL/s with the spray noz-zle heated at 60 ◦C allowed the rectangular application of 250 �Lcheese sample within 6 or 4 min, respectively. As this step wascompletely automated, the application time was acceptable. Forroutine use, 1-�L application volumes were sufficient for controlof the strictest lactose level discussed so far. At a dosage speed of250 nL/s, this 1-�L application took only 4 s.

3.7. Evaluation of the working range

The working range started with the LOQ at 3 ng/band. Theregression analysis was performed over a wide 1:100 concentra-tion range between 3 and 300 ng/band. The regression analysesof lactose and galactose were slightly polynomial with correla-tion coefficients r of ≥0.9999, both for peak height and area (Fig.S-3). The relative standard deviations of the four calibration curveswere ≤2.3%. The performance data proved the very good corre-lation between signal and concentration over the wide workingrange of 1:100 for lactose and galactose. For routine use, a nar-rower 1:15 calibration range (3 to 60 ng/band) was sufficient, alsowith a correlation coefficient r of 0.9999. Calibrations in the range10–150 ng/band showed correlation coefficients r of 0.9998, 0.9999and 0.9996.

3.8. Comparison of the calibration with and without matrix

For investigation of the matrix influence on quantitation, the cal-ibration was performed in 1 and 10 �L cheese matrix, previouslyproven to be ‘lactose-free’ at these application volumes. Five lactosestandard levels ranged 3–120 ng/band were applied both solely andover-sprayed each on a 1 �L and 10 �L sample band. The calibrationcurves without and with matrix for a 1 and 10 �L sample loadingwere compared and found to be equivalent (Fig. 2). Hence, an exter-nal standard calibration can be used for control of the lactose levelsin routine use. However, at a 5-times higher matrix load, e.g. 50 �L,the calibration curve was shifted by the intrinsic lactose level of thecheese matrix itself. As already mentioned, no blank sample wasavailable for such low LODs and in all samples lactose was detectedat increased application volumes.

3.9. Determination of recovery rates

The recovery was performed via standard addition of lactose to

a wide range of dairy products spiked at the level of the strictestlactose value discussed. Aliquots of the aqueous lactose solutionwere added to the sample (100 mg/L or 0.01%), mixed and extracted.The spiked sample and the original sample were two-fold applied

220 G.E. Morlock et al. / J. Chromatogr. A 1324 (2014) 215– 223

F1

earpm9o

3

sapuchwtmotar

iAa

TRs

ig. 2. Comparison of the external calibration curve with the calibration in 1 and0 �L cheese matrix showing determination coefficients R2 ranged 0.9928–0.9999.

ach (Fig. 3). The original lactose content of samples was subtractednd then the recovery rate was calculated. The mean results (n = 2)anged between 78.9 and 105.3% over the broad range of dairyroduct types with precisions between 0.2 and 2.9% (Table 4). Theean recovery over all dairy products investigated (n = 11) was

0.5 ± 10.5% and the mean precision of the two-fold determinationf all recovery rates was 1.3 ± 1.0%.

.10. Proof of matrix interference

HPTLC easily offers options to gain further information on theample matrix. As it is described that amino compounds and sug-rs can react according to a Maillard reaction to blue fluorescentroducts through heating on a plate [22,23], diverse dairy prod-ct matrices (yoghurt no. 2, milk no. 3, sour cream no. 7 andream cheese no. 16) were selected and derivatized with the nin-ydrin reagent to prove for interferences (co-eluting compoundsith amino groups), which could reveal during the derivatiza-

ion (heating) and potentially increase the results. At the lactoseigration distance of 21 mm or hRF 22, no amino compounds were

bserved, even after prolonged heating time (15 min) at increasedemperature (140 ◦C). Hence, for the different matrices investigatedny interference with matrix components was not observed withegard to detection and the selectivity was confirmed.

A potential hR shift of the lactose zone due to matrix

F

nterference was investigated by overlapped application (Fig. 4).pplication was performed so that 16-mm bands each of a samplend standard solution overlap by 50%. Using this technique, one

able 4ecovery rate calculated via standard addition of lactose to various dairy productspiked at the strictest lactose value discussed (100 mg/L).

Lactose-freedairy product

No. Mean recoveryrate of lactose (%)

Repeatability(%RSD, n = 2)

Butter 1 105.3 1.9Yoghurt 2 87.7 0.7Milk 3 87.7 0.6Evaporated milk 5 76.2 3.1Buttermilk 6 81.2 0.2Sour cream 7 78.9 2.9Cream 8 82.8 0.2Goat cheese 9 103.6 1.4Cheese 15 102.4 1.2Cream cheese 16 99.2 0.5

17 90.3 1.8Mean ± %RSD (n = 11)

for all dairyproducts

90.5 ± 10.5 1.3 ± 1.0

Fig. 3. Two-fold determination of the recovery rate via standard addition of lactose(Lac) to various milk products at the strictest lactose limit value discussed (0.01%)

and of the original sample (no. assigned in Table 2, standard levels S1-S4) on platesA–C (UV 366 nm) with respective 3-D densitograms (366/>400 nm); glucose (Glc)and galactose (Gal) are visible in samples as well.

part of the lactose standard migrated in the sample matrix, andhence, any hRF shift was easily discovered and proven. In the goatcheese (no. 10) lactose was visible and confirmed by overspray withthe lactose standard. In the other cheese (no. 12) lactose was notfound. Evaluation at UV 366 nm was more sensitive than inspectionin the visible light.

After chromatography, the separation was inspected underwhite light illumination for visible interferences, under UV 254 nm

for UV absorbing interferences (in case of using plate with F254 indi-cator) and under UV 366 nm for natively fluorescent compounds in‘lactose-free’ labelled dairy products. For higher application vol-umes (e.g. 10 �L), many fluorescent compounds were visible at

G.E. Morlock et al. / J. Chromatogr. A 1324 (2014) 215– 223 221

Fsm

U(mfltb

3

talaa1ltaaiwi8av‘rttd

3o

cw(aEaosbpwf2p

Fig. 5. Quantitation of samples: 3D-densitogram with lactose standard levels (S1-S5, 3–60 ng/band) and dairy products (n = 3, 1 �L applied as 8 mm band each): milkno. 3 and evaporated milk no. 5 contained 0.007% and 0.006% lactose, respectively

ig. 4. Overlapped application of cheese samples (no. 12 and 10) and lactosetandard (Lac) for proof of any hRF value shift caused by matrix interference, docu-ented at white light illumination and UV 366 nm.

V 366 nm, however, not at the later migration position of lactosedotted line in Fig. S-4), which is not directly visible after chro-

atography. This method could also be used for analysis of theseuorescent compounds, which were separated well. Hence, usinghis technique multi-fold information on the samples could quicklye gained and assist the analyst in understanding matrix problems.

.11. Sample analysis

Commercially available ‘lactose-free’ dairy products were inves-igated. Volumes of 1 �L of the 10% dairy product extracts werepplied to control the LOD level of 0.001%, which is the strictestactose level discussed so far. Three out of the 17 commerciallyvailable ‘lactose-free’ labelled dairy products contained lactosebove the selected level of 0.001% (Table 2). In the goat cheese (no.0) 0.064% lactose was detected. In the ‘lactose-free’ milk (no. 3)

actose was determined to be 0.007%. In evaporated milk (no. 5)he content found was 0.006% (Fig. 5A). All samples were concord-nt with the label (<0.1 g/100 g), but the cheese sample no. 10 wasbove the strictest limit value of 0.01% discussed by the LChG work-ng group. In each ‘lactose-free’ labeled sample lactose was detected

hen higher application volumes were chosen (Fig. 5B; already vis-ble for 80 �L butter (no. 1) and 10 �L yoghurt (no. 2) or cream (no.)). With regard to the galactose content, the cheese samples (no. 9nd 15 in Fig. 3) did not contain galactose above the strictest limitalue of 0.01% discussed, in contrast to the enzymatically degraded

lactose-free’ milk products. The butter (no. 1) contained compa-ably less galactose than milk and milk-products, however, morehan cheese. Hence, for galactosaemia patients only cheese (provedo be lactose-free) could be recommended among the ‘lactose-free’airy products investigated.

.12. Analysis of costs and analysis time—Benchmarking withther methods

For routine analysis and control of the strictest lactose level dis-ussed, the application of 1-�L sample volumes on 8 mm bandsas used (Fig. 6). This allowed the parallel screening of 20 samples

tracks) or the quantitation of 16 samples (plus 4 standard levelspplied). Costs for the HPTLC plate were about 5.40 Euro. About 0.5uro were calculated for solvent consumption (10 mL per plate)nd its disposal. The derivatization reagent was re-used at leastver two months and about 1.5 mL was consumed by each immer-ion step. Thus the costs for the derivatization were calculated to beelow 0.1 Euro. All in all, the HPTLC running costs were about 6 Euroer plate. For 20 analyses per plate, the running costs per analysis

ere 0.3 Euro (or 0.4 USD). If compared to current HPLC methods

or lactose determination in milk products with gradient times of4 to 65 min at flow rate of 0.4 to 1 mL/min [14,15], the HPLC mobilehase consumption per analysis on its own is substantially higher.

(A); lactose was found in all samples (e.g., butter no. 1: 80 �L; yoghurt no. 2 andcream no. 8: 10 �L each) when higher volumes were applied as 15-mm bands (B).

Moreover the stationary phase has to be protected by the use ofpre-columns and an extended sample preparation. Column chro-matographic methods might have problems in routine use whena reduced sample preparation is applied or varying matrices areanalyzed over many separations on the same column. Universaldetectors used for carbohydrate analysis by HPLC are not selectiveand also require extended sample preparation for routine use. Whytake so much effort?

Hence, the simplest approach was chosen to quantify lactose indairy products. This streamlined analysis of ‘lactose-free’ food usesapplication volumes of up to 250 �L for a 10% sample solution andthe high matrix tolerance is evident. The stationary phase is alwaysfresh and a matrix-carry over is not in question and definitely notinterfering with subsequent runs. Thus, the main advantage wasthe high method robustness despite the high matrix loading onthe adsorbent. If compared to column-based methods, such a highmatrix loading at the given simple sample preparation is impossi-ble (large-volume liquid injection of 250 �L of a 10% concentratedcheese extract) and solid phase extractions and/or Carrez I and IIprecipitations are included in the protocol.

All HPTLC steps were performed automatically and controlledin a common software file. Only a transfer time of 5 min wasrequired between the instruments for 20 tracks per plate. The timeper analysis was calculated to be 3 min when 20 analyses wereperformed on a plate with regard to a LOD level of 0.001% lactose.

It took 18 min for application (16 1-�L samples and 4 standardlevels), 30 min for separation, 7 min for derivatization and 5 minfor densitometry, which summed up to 60 min per plate containing

222 G.E. Morlock et al. / J. Chromatogr. A 1324 (2014) 215– 223

F n of da

2oiaotsl6owsu

4

tavcrUaiaatd

ofuwipat

[

[

ig. 6. Sample screening illustrated for 20 parallel analyses: two-fold determinatios well as standard levels (S1-S4, 1–100 ng/band).

0 tracks. HPTLC is an offline method and the single instrumentsf a HPTLC working station are automated. As it is automated ints single steps, one plate can be applied, a second be developednd a third is measured–all at the same time. If the instrumentsf the HPTLC workings station are used in a 30-min interval shift,he sample throughput can be doubled and is calculated to be 300amples per 8-h day. If compared to current HPLC methods foractose determination in milk products, gradient times of 24 to5 min were reported [14,15]. This enables a sample throughputf 22 to 60 samples per 24-h day. However in general, more effortas spent on sample preparation, and fortunately, this was kept

imple when HPTLC was used. These considerations make clear thenrivaled potential of HPTLC for food screening and food safety.

. Conclusions

The newly developed HPTLC method provided a high samplehroughput screening of dairy products, and at the same time,

reliable quantitative analysis of any future ‘lactose-free’ limitalue, inclusive of its correct labelling with regard to the galactoseontent. Considering the one-step extraction, good detectability,obustness with regard to varying matrix, the low running costs (0.5SD/analysis) and the fast analysis time (3 min) at the 10 mg/L levels well as the good performance data discussed, the HPTLC methods recommended for a cost-effective routine analysis in food safetynd quality control of dairy products. The developed method willlso support research and development of more advanced produc-ion technology for lactose-free dairy products due to its very lowetectability of lactose.

Although HPTLC is underestimated by food analysts, the meritsf the described HPTLC method for lactose analysis in ‘lactose-free’ood products are evident. Keeping all these aspects in mind, it is notnderstandable that planar chromatography was not consideredhen research projects are funded and financed that deal with the

ssue of selecting proper analytical methods for ‘lactose-free’ foodroducts [24]. HPTLC changed over the last decade and became anttractive option for analysts [25], as it would be advantageous tohem to select the most suited method to solve a given task. To

[

ifferent milk products (1 �L applied each as 8-mm bands, no. assigned in Table 2)

improve the low visibleness of HPTLC for analysts, we have to givespace to proper training in HPTLC at educational sites.

5. Author contributions

CL performed selection of the mobile phase and some initialtests. LPM performed all other experiments in the framework of apractical orientation course for scholars (BOGY). Both were super-vised by GEM.

Appendix A. Supplementary data

Supplementary data associated with this article can be found,in the online version, at http://dx.doi.org/10.1016/j.chroma.2013.11.038.

References

[1] G.E. Morlock, W. Schwack, J. Planar Chromatogr. 20 (2007) 399–406.[2] S.W. Souci, W. Fachmann, H. Kraut, Food Composition and Nutrition Tables,

Seventh ed., CRC Press Inc., Boca Raton, 2008.[3] EFSA Panel on Dietetic Products, EFSA Journal 8 (9) (2010) 1777 (29 pages).[4] Working group Issues of Nutrition, Lebensmittelchemie 59 (2005) 45 (2 pages).[5] German Milk Industry Association, Background Information on Lactose, in:

Paper of the German Milk Industry Association, German Milk Industry Asso-ciation, Berlin, March 2011 (4 pages).

[6] The Council of the European Union, Council Regulation (EC) No 1234/2007 of22 October 2007 Establishing a Common Organisation of Agricultural Mar-kets and on Specific Provisions for Certain Agricultural Products (Single CMORegulation), Annex XIII, III, 2 c.(former EG VO 2597/97, 2007.

[7] Gesetz über Milch, Milcherzeugnisse, Margarineerzeugnisse und ähnlicheErzeugnisse (Milch- und Margarinegesetz - MilchMargG), §4 Zulassung vonAusnahmen, 25.07.1990, BGBl. I S. 1471, Artikel 22, 09.12.2010 BGBl. I S,1934.

[8] J. Sahlstein, O. Tossavainen, Valio Ltd., Process for Producing a Lactose-Free MilkProduct, WO 2003094623 A1, 20.11.2003.

[9] AOAC, AOAC Official Method 984.15 Lactose in Milk, Enzymatic Method, AOAC,1985.

10] J.A. Sanchez-Manzanares, M.R. Fernandez-Villacanas, F. Marin-Iniesta, J.Laencina, Food Chem. 46 (1993) 425–427.

11] DIN 10344, Milk and Milk Products—Determination of Lactose and d-galactose

Content—Enzymatic Method, Beuth Verlag, Berlin, 2013.

12] IDF-Standards 79-1 and 79-2, ISO 5765-1:2002 and ISO 5765-2:2002, DriedMilk, Dried Ice-Mixes and Processed Cheese—Determination of LactoseContent—Part 1 and 2: Enzymatic Method Utilizing the Glucose Moiety of theLactose, 2002.

matog

[

[

[

[[

[

[[

[[[[

G.E. Morlock et al. / J. Chro

13] ISO 22662:2007 (IDF 198: 2007), Milk and Milk Products—Determinationof Lactose Content by High-Performance Liquid Chromatography (ReferenceMethod), Beuth-Verlag, Berlin, 2008.

14] ESA, Dionex, Application Note: Lactose in milk—A Validated Method, 2011 (yearnot available).

15] Thermo, Application Note 248: Determination of Lactose in Lactose-Free MilkProducts by High-Performance Anion-Exchange Chromatography with Pulsed

Amperometric Detection, 2010.

16] X. Zhang, Y. Cao, J. Ye, Food Chem. 72 (2001) 385–388.17] G. Morlock, G. Sabir, J. Liq. Chromatogr. Related Technol. 34 (2011)

902–919.18] G.E. Morlock, C. Oellig, J. AOAC Int. 92 (2009) 745–756. [

r. A 1324 (2014) 215– 223 223

19] G.E. Morlock, S. Prabha, J. Agric. Food Chem. 55 (2007) 7217–7223.20] L.W. Kroh, M. Zietz, Y. Pfeifer, M.-A. Bornik, S. Rohn, Dtsch. Lebensm. Rundsch.

April (2010) 219–223.21] M. Pukl, M. Prosek, J. Planar Chromatogr. 3 (1990) 173–176.22] R. Klaus, W. Fischer, H.E. Hauck, Chromatographia 28 (1989) 364–366.23] R. Klaus, W. Fischer, H.E. Hauck, Chromatographia 29 (1990) 467–472.24] State Institute for Teaching and Research on Milk—Dr. Oscar-Farny Insti-

tute (LAZBW), Research project no. 06HS012 “Lowering the DetectionLimit in the Determination of Lactose and Galactose in ‘lactose-free’ DairyProducts”, 2007–2008, http://www.fisaonline.de/index.php?act=projects&p id=768&lang=en. (90 pages).

25] G. Morlock, G.I.T. Lab. J. 5–6 (2012) 12–13.